Checkpoints

Increasing the value of this parameter speeds up the export of big data. Set this parameter based on checkpoint_timeout and shared_buffers. This parameter affects the number of WAL log segment files that can be reused. Generally, the maximum number of reused files in the pg_xlog folder is twice the number of checkpoint segments. The reused files are not deleted and are renamed to the WAL log segment files which will be later used.

If the value of checkpoint_segments is increased, you need to increase the value of this parameter. The increase of them further requires the increase of shared_buffers. Consider all these parameters during setting.

• When delayed checkpoint deletion is enabled, the primary DN will move the relfilenode file to be deleted to the pg_delayed_unlinks_bin directory of the corresponding database when creating a checkpoint.
• To determine whether an OID is reused, the system not only checks whether the relfilenode file exists in the corresponding database directory, but also checks whether the pg_delayed_unlinks_bin directory exists.
• After the standby DN replays to the corresponding checkpoint record, the primary DN will delete the batch of relfilenode files in the pg_delayed_unlinks_bin directory corresponding to the checkpoint replayed by the standby DN in the next checkpoint.
• When there are enough delayed-deletion relfilenode records (10 million), the excess relfilenode files will remain in the pg_delayed_unlinks_bin directory until the corresponding database/tablespace is deleted or the node is restarted and the files are cleaned up.